When Park High School in North London gained an extra 600 students and began the move into a virtualised environment, network manager David Crawley knew the school would need network and application monitoring tools to support growth.
Over time, Park High's network has developed piecemeal, connecting a total of 1,700 students and 170 staff, with support for a mixture of laptops, desktops, thin clients, PCs, Macs and a partially virtualised server estate.
“As with most schools, it’s very much something that has grown sporadically over the years, and we do tend to run computers a year or two longer than ideally you’d like to. So it can be quite a challenging network to manage,” said Crawley.
In the past, attempts to monitor the network were “probably a little backwards,” he added. “If you had a device that was slowing down your entire network, to track that back you’d pretty much have to start at the centre of your network and work out step by step, he said.
“[You would] log on to every switch individually, see if there was a particular port that was having too much traffic. If the traffic was too bad, you may not even be able to log on to a switch; you’d end up having to take a laptop, plug a console cable in and query it directly.”
Old-school troubleshooting is especially lacking considering the strain that virtualisation can place on networks. “[With virtualisation] one of the key aims is to get more processing power out of each physical server by loading it with as many virtual machines as is reasonable. That stresses the network and the network bandwidth, and network I/O assigned to a physical server may become the main performance bottleneck, rather than CPU and storage,” said Quocirca analyst Bob Tarzey.
So Crawley sought a network monitoring tool that could handle both physical networks and virtual environments, choosing to pilot the PRTG Network Monitor from Paessler, which provides a real-time view of network activity as well as historical reporting.
The monitoring tool would start by taking on the physical network, which is based on a 10 Gb, 12 core fibre backbone with 1 Gb fibre links and HP ProCurve switches. This summer, Crawley's team also installed a new wireless system based on Netgear wireless routers with a Wavesight wireless bridge that went live at the start of the new term in September 2011.
Application monitoring in the network
After first focusing on basic network functions, Crawley will extend the reach of the monitoring tool to look more deeply at how applications are performing. “We are in [the] process of updating our school’s website, and once the new system is in place, we'll monitor the external Web server,” said Crawley.
Eventually that monitoring will drill down into some applications to look at specific transactions. “We do have the ability to monitor something that runs on a server as a service. I can monitor their uptime and make sure that they're not exceeding any threshold and that they're just essentially working for people.”
Yet some applications are still beyond the reach of most network monitoring tools. “It would be useful if you could monitor certain applications, but a lot of that comes down to the application developers themselves making that information available to external monitoring solutions to read. That interoperability isn’t always there,” he said.
Still, applications that are available for monitoring can be used as a gauge for how well networks are working and vice versa, said Tarzey. “When it comes to monitoring application performance, there are aspects you can report on which do not need much access to the application itself; for example, end-to-end performance ... if you can establish that the network is working well, then that is the time to take a closer look at the application, however you achieve that,” he said.
Network monitoring in a virtual environment
The monitoring solution also allows Crawley to keep a weather eye on the virtualised estate and pre-empt any bottlenecks, as well as avoid overprovisioning. Yet this is mostly a manual operation. “It’s very useful to see if one of my physical hosts is particularly overloaded for the task, [so I can] migrate certain virtual servers to other hosts,” said Crawley.
Using the PRTG monitoring tool also enables Crawley to avoid overprovisioning. “The PRTG monitoring has been helpful if I’m taking services or servers that are currently running on physical hosts and determining just how much processor capacity and RAM they actually needed to give to the virtual machine. It’s meant that we can then not need to overprovision our virtual servers too much and make more use of what we’ve got.”
What's more, remote monitoring of the virtual environment and network proved invaluable during a recent outage. “I was on holiday and one of our fans went offline,” Crawley said. “This took out about two-thirds of our virtual machines, and at the moment the majority of our servers are now running on virtual machines, so that essentially took out the majority of the network.
“I was able to get on to the network remotely, see exactly which fan had gone down, which physical hosts it had taken down, which virtual machines were down. I was able to get the network back up and running from home within 45 minutes.”